fix: remove GPU-dependent pre-download, use restructure_pages for cross-page tables, robust md extraction

fix: use CPU mode for model pre-download during Docker build

feat: v5.0.0 PaddleOCR-VL-1.5 + Gemini hybrid architecture

feat: v4.0.0 — VLM + Gemini 3 Flash hybrid (table pages use Gemini API)

feat: v3.3.1 - disable table re-prompting, add page number cleanup

feat: v3.3.0 - table re-prompting, heading normalization, footer cleanup

feat: v3.3.0 - increase max_tokens to 32768 for wide tables

feat: v3.3.0 - heading normalization, footer cleanup, table fixes

feat: v3.3.0 - DPI 200, post-processing, cross-page dedup

fix: total_mem → total_memory attribute fix for startup

feat: v3.2.1 - remove page markers, fix escaped quotes

feat: v3.2.0 - LaTeX→MD conversion, VLM output cleanup, improved prompt, disable thinking

feat: v3.1.0 - DPI 150, parallel rendering, VLM retry, quality fixes

feat: v3.0.0 VLM-first hybrid architecture — GPU VLM on all pages, Docling TableFormer only on table pages

fix: reduce concurrent VLM workers to 2 to prevent GPU OOM on 30B model

perf: concurrent VLM OCR — process pages in parallel via ThreadPoolExecutor

fix: resolve /parse/url for URLs without file extensions (e.g. arxiv)

feat: increase VLM max_tokens to 16384

fix: increase max-model-len to 65536 for VLM image tokens, improve error logging

fix: reduce max_tokens to 4096, remove invalid skip_special_tokens, add error body logging

fix: add onnxruntime for Docling RapidOCR, fix total_memory attr

fix: total_mem -> total_memory, clean up debug CMD and start.sh

debug: use python3 CMD to test if output reaches HF logs

debug: inline bash -c to test if CMD even executes

fix: use full path /bin/bash entrypoint to ensure CMD executes

debug: add aggressive startup logging to diagnose silent hang

fix: remove sed pipe that caused block buffering and wrong PID tracking

fix: use HF repo ID for vLLM model loading instead of local path

fix: use JSON format for --limit-mm-per-prompt (vLLM compat)

fix: use python3 in start.sh (vLLM base image)

feat: upgrade to Qwen3-VL-30B-A3B, simplify auth, fix redirects

fix: remove CLAUDE.md from HF Space (internal dev file)

docs: fix deploy instructions for monorepo, remove numpy dep

feat: hybrid VLM parser with Qwen3-VL-8B via vLLM (v2.0.0)

feat: support both API_TOKEN and API_DEV_TOKEN

fix: single-line Python command for HF Docker builder

Initial commit: Docling Parser API

Initial commit: Docling Parser API